Aims. We selected a large sample of O-B stars that were considered as (candidate) slowly pulsating B, ? Cep, and Maia stars after the analysis of their hipparcos data. We analysed our new seven passband geneva data collected for these stars during the first three years of scientific operations of the mercator telescope. We performed a frequency analysis for 28 targets with more than 50 high-quality measurements to improve their variability classification. For the pulsating stars, we tried both to identify the modes and to search for rotationally split modes. Methods: We searched for frequencies in all the geneva passbands and colours by using two independent frequency analysis methods and we applied a 3.6 S/N-level criterion to locate the significant peaks in the periodograms. The modes were identified by applying the method of photometric amplitudes for which we calculated a large, homogeneous grid of equilibrium models to perform a pulsational stability analysis. When both the radius and the projected rotational velocity of an object are known, we determined a lower limit for the rotation frequency to estimate the expected frequency spacings in rotationally split pulsation modes. Results: We detected 61 frequencies, among which 33 are new. We classified 21 objects as pulsating variables (7 new confirmed pulsating stars, including 2 hybrid ? Cep/SPB stars), 6 as non-pulsating variables (binaries or spotted stars), and 1 as photometrically constant. All the Maia candidates were reclassified into other variability classes. We performed mode identification for the pulsating variables for the first time. The most probable l value is 0, 1, 2, and 4 for 1, 31, 9, and 5 modes, respectively, including only 4 unambiguous identifications. For 7 stars we cannot rule out that some of the observed frequencies belong to the same rotationally split mode. For 4 targets we may begin to resolve close frequency multiplets.